We used a modified light weight video restoration transformer that predicts intermediate frames without any optical flow estimation by looking only the previous frame or following causality.
Dataset: DAVIS
Checkpoints: checkpoints
Presentation: Presentation